►
From YouTube: Kubernetes - AWS Provider - Meeting 20220819
Description
Recording of the AWS Provider subproject meeting held on 20220819
Agenda - https://docs.google.com/document/d/1-i0xQidlXnFEP9fXHWkBxqySkXwJnrGJP9OGyP2_P14/
A
Good
morning,
folks,
today
is
the
august
19th
2022.
welcome
to
the
sig
it
cloud
providers,
aws
subproject
meeting,
please
be
respectful
with
your
communication
and.
B
A
To
the
cncf
code
of
conduct,
I
am
kishi
your
host
for
today.
So,
let's
add
the
agenda
for
today's
meeting
here.
C
I
guess
my
update
that
I
put
in
the
generic
agenda
yeah.
We
just
need
a
new
build
for
123,
at
least
and
I'll
have
to
see
what
other
versions.
So
I'm
planning
on
doing
that.
A
C
Yeah
it'll
be
issue
and
also
pratik
has
the
sts.
A
I
see
okay,
I've
seen
that
pull
pr
as
well:
okay,
cool
k,
ops,
justin.
D
Yeah,
nothing
particularly
specific
this
week
of
general
interest,
though,
is
this
run
c
regression
around
executive
tty,
which
we
are
grappling
with
for
124.
I
will
try
to
find
a
link
and
paste
it
into
the
notes
for
anyone
that
is
interested
or
may
have
a
solution
other
than
asking
run
c
nicely
to
do
a
new
release.
I
think
they're,
aware
of
it
so.
A
Cool,
thank
you.
The
next
load
balancer
control,
so
we
released
the
v243
patch
release,
nothing
major
changes,
but
we
added
some
error:
handling
more
resiliency
on
certain
cases
and
bunch
of
community
contribution
out
there.
243
is
out.
C
Yeah,
do
you
know
who
can
we
invite
that
would
be
willing
to
like
his
code.
E
E
Team
we
have
a
few
folks.
If
we
want
somebody
from
pm
side,
it
would
be
alex
or
engineering
side
would
be
either
alice,
brandon
nick.
So
those
guys
you
can
think
of
okay.
E
C
Okay,
yeah
I'll,
maybe
either
me
or
you-
can
extend
the
invite
and
see
if
they
want
to
send
someone.
C
See
there
I
I
did
see
something
I
saw
sabrina
had
given
the
lgtm.
I
I
got
some
things
this
morning.
I
I'm,
like
tagging
controller,
had
a
bug.
Maybe
we
should
just
look
at
that.
You
want
to
open
up
the
ccm
really
quick
and
see.
A
A
A
C
A
C
Yeah,
I
have
no
idea,
that's
a
great
question.
I
wonder
if
sora
has
opened
an
issue
upstream
as
well,
it
might
be
worth
asking
see.
A
Yeah,
it
doesn't
look
any
complicated
to
me.
I
will
also
check
like
whether
it's
ccm
specific
or
how
we
have
been
doing
in
the
entry,
the
kcm
earlier,
we're
not
going
to
backport
it
or
anything
it.
It
will
likely
merge
this
and
move
on.
D
Oh
sorry,
the
one
concern
I
have
is
it
sounds
like
this
was
a
specific
problem
with
nodes
and
that
this
tag
resource
is
called
from
a
bunch
of
places
and
the
one
thing
I
think
we
would
want
to
be
careful
of
is,
for
example,
if
there
was
an
async
or
eventually
consistent
api.
Like
you
know,
iam
sometimes
doesn't
show
things
right
away
or
did
at
least
used
to
not
show
things
right
away.
D
We
would
probably
lose
the
ability
to
retry,
maybe
so
the
the
alternative
way
to
do
this
would
be
for
the
specific
case
when
we're
calling
tag,
resources
or
untech
resources
to
check
for
a
check
for
this
particular
error
on
the
call
site.
But
I
do
kind
of
agree
that
I
think
this
is
probably
pretty
a
pretty
good
idea
like.
I
think
this
tagging
is
pretty
much
best
effort
anyway,
but
it's
worth
checking
where
it
is
cold.
A
C
B
C
I
don't
have
anything
in
mind.
Okay,
I
guess
yeah,
just
like
me
and
pratik
are,
I
think,
looking
at
the
cherry
picks
that
he
opened
they're
failing
on
some
like
the
ede
tests
are
failing
or
something
so
we
need
to
figure
out.
What's
what's
going
on
there
cool.
A
I
have
a
question
for
justin
earlier,
like
john
from
chaos.
Team
was
talking
about
like
how
we
want
to
include
the
load
balancer
controller
with
chaops
as
a
add-on,
and
he
also
brought
about
like
disabling
the
cloud
provider,
the
load
balancer
code
in
the
ccm.
So
that
is
something
that
I
or
we
want
to
look
into
as
well.
A
So
we
want
to
also
have
like
the
load
balancer
controller
as
like
a
managed
add-on,
which
is
available
by
default.
So
any
thoughts
on
that
from
your
site
from
the
k,
ops,.
D
Yeah,
I
mean
you're,
I
think
I
think,
nick
you've,
you
integrated
the
cloud
provider
aws
in
the
past,
so
you
probably
have
the
most
like
direct
experience
of
of
someone
like
adding
something
there
you'd
be
very
welcome
to
that
would
be
wonderful.
We
also
are
working
on
more
what
we
call
sort
of
additional
objects
or
external
objects.
D
So
the
idea
that
if
the
load
balancer
controller
had
had
a
something
like
component
config,
you
could
sort
of
configure
it
that
way
without
having
to
necessarily
integrate
it
as
tightly
into
the
k-ops
api
or
into
the
one
day
or
as
tightly
into
the
k-ups
code.
We
aren't
totally
there
yet
and
that's
still
future
flagged,
so
we're
sort
of
in
between
the
two
worlds.
Okay,
but
you
know,
if
you'd
be
welcome
to
do
either
one
and
I'm
happy
to
sort
of
give.
D
You
pointers
if,
if
need
be,
like
more
detailed
pointers,
if
need
be
so.
A
Component
company
is
like
a
kia
specific
resource.
D
Now
so
I
mean
component
config
is
this:
is
this
idea
in
upstream
kubernetes
of
replacing
flags
with
something
which
looks
like
a
crd?
It's
actually
a
file,
but
it
is,
I
think
it
uses
the
the
api
machinery.
So
it
is
a
it
is
a
typed
and
versioned
object.
D
It's
sort
of
like
scheduler,
moved
to
it
fairly.
We
had
a.
We
had
an
issue
in
k,
ups,
where
scheduler,
basically
deprecated
a
function,
functionality
from
flags
and
you
had
to
start
using
the
scheduler
config,
yaml
or
file,
and
that
was
unfortunate
because
we
didn't
we
didn't
transition
in
time
for
some
people.
But
yes,
it's
it's.
Basically
a
gamma
configuration
file
instead
of
flex.
A
Cool
sure
I
will
definitely
look
into
it
because
that's
the
the
direction
that
we
want
to
head
into
as
well,
because
the
like
the
load
balancer
code
in
the
cloud
file
radar.
We
haven't
been
updating
it
recently.
A
So
we
want
to
replace
that
particular
component
with
the
lbc,
because
that's
where
most
of
the
active
development
takes
place
for
us
yeah
cool.
I.
D
A
See
nothing
in
the
immediate
roadmap
right
for
the
kiops.
D
A
C
On
that
note,
do
we
want
to
pack
it
like,
I
feel
like
we
want
most
cases,
users
to
install
ccm
and
the
load
bouncer
controller
together.
So,
like
you
know,
cops
is
one
situation.
Do
we
also
want
to,
like
I
don't
know,
somehow,
have
a
packaged
installation
method
for
for
users
who
are
not
using
eks
and
not
using
cops.
A
So
we'll
provide
the
options,
that's
the
best
thing.
We
can
hope
right,
because
different
distributions
have
their
own.
Our
packaging
have
their
own
rules
and
requirements
will
will
enable
them,
but
will
not
go
and
more
define,
like
all
the
possible
packages
out
there.
Maybe
the
popular
ones.
Chaos
cks
like
these
are
the
ones
that
that
are
more
popular,
we'll
do
something
there
for
other
users,
we'll
let
it
up
to
the
maintainers
to
pick
it
up.
At
least
we
will
have
the
option
so
we'll
say:
okay,
this
is
the
one.
C
Yeah,
I'm
thinking
of
users
who
come
to,
I
don't
know,
maybe
they're
you
know
they
have
some
way
of
setting
a
cluster.
They
come
to
the
ccm
repository
and
they
install
it
somehow
right
now,
there's
not
a
lot
of
information
about
the
load,
balancer
controller
from
that
direction,
so
maybe
even
even
as
just
like
documentation,
we
just
add
good
documentation
on
lbc.
A
Sure
so,
once
we
provide
the
flag
right
so
then
we
can
make
it
disable
it
by
default.
That
way,
like
users
will
be
forced
to
do
something
about
it.
Right
either
go
ahead
and
enable
the
flag
or
install
the
lbc.
So
at.
A
That
we
can
provoke
from
the
user
side.
That's
my
thought
process
right
now.
Then
that
point
we
can
definitely
like
start
adding
the
documentation,
yeah.